Study On Noise Removal And Sea Ice Cover Detection Methods Based On Sentinel-1 SAR Cross-polarization Extra-wide Swath Data

Sentinel-1 constellation,as the C-band synthetic aperture radar(SAR)satellite,can provide cross-polarized(horizontal-vertical,HV or vertical-horizontal,VH)images in extra-wide swath mode(EW,up to 400 km)in high spatial and temporal resolutions,to monitor sea ice in polar regions.The cross-polarized SAR data has low sensitivity to radar incidence angles and sea surface roughness,making it more suitable for sea ice detection.However,cross-polarized EW mode SAR data are significantly disturbed by thermal noise due to low signal to noise ratio.To better use Sentinel-1 data for sea ice monitoring research in the Arctic,especially for sea ice detection in the marginal ice zone(MIZ),the thermal noise removal method is studied based on the Sentinel-1 HVpolarized EW data from January 2017 to December 2019.Based on the de-noised Sentinel-1 data,the automatic sea ice detection method is further studied.The main contributions of this study are in following.(1)The study of the thermal noise removal method for Sentinel-1 HV-polarized EW images.Based on the thermal noise removal method proposed by Park et al.,(2018),the calculation of the optimal scaling factor and balance factor of the range noise vectors(NESZ vectors)is improved,to better correct the 2-D additive noise field;Then an algorithm based on the same NESZ vectors is proposed to eliminate the multiplicative noise by correcting the signal's backscattering amplitude to be uniform along the whole range direction.Examples verification indicates that the denoising algorithm proposed in this study can effectively remove the thermal noise in the Sentinel-1 HV-polarized EW images and improving the image quality,achieving better results than the Park(2018)method.This is beneficial to use the denoised data to carry out studies on sea ice montoring in the Arctic.(2)Using the denoised Sentinel-1 HV-polarized EW images,the previously developed sea ice cover detection method by our research group for Gao Fen-3 SAR images is further improved.The improvement includes the image preprocessing,automatic extraction and classification of training samples,and the window parameters in texture calculation.As a result,the improved method is effectively applied to the Sentinel-1 HV-polarized EW data to obtain sea ice cover data in high resolution in the Arctic MIZ.(3)Compared with visual interpretation of cases,the accuracy of sea ice cover data obtained by the above method is more than 95%.Compared with with the IMS(Interactive Multisensor Snow and Ice Mapping System)products based on 728 scenes of EW data,the average accuracy of sea ice coverage data can reach approximately 80%.We also tried this method to the X-band SAR data of Terra SAR-X for a few cases,which also suggests good results.